拉延筋等效模型的标定及其多参数映射神经网络  

Calibration and multi-parameter neural network of the equivalent model of drawbead

作　　者：孙德林[1] 曾攀[1] 

机构地区：[1]清华大学机械工程系,100084

出　　处：《锻压技术》2004年第1期21-26,共6页Forging & Stamping Technology

基　　金：国家杰出青年基金资助 (5982 51 1 7)

摘　　要：拉延筋在实际板料成形工艺中可以控制材料的流动 ,有着极其重要的作用。由于拉延筋的结构复杂 ,相对尺寸较小 ,影响的参数多 ,表现出高度的局部非线性 ,对数值模拟中的计算精度和效率都有重要的影响 ,因而 ,在工程模型中需要建立相应的等效计算模型 ,这就要进行等效标定。本文首先建立起用于等效标定的计算模型 ,研究并提出了基于金属流动速度的等效“标定准则” ,给出了包括工艺和几何参数影响的标定关系 ,基于神经网络 ,用单参数离散标定关系来构建多参数连续标定关系 ,最后 ,进行方盒冲压过程的实际模拟及标定验证。Drawbead is used to control the flow of the material in sheet metal forming processes. Because of the significant effects of the complicated structure, relative small size, and multiple highly local non\|linear parameters of drawbeads to the precision and efficiency of the simulation, the equivalent model plays an important role. Therefore, the calibration is needed.In this paper, the method for calibration is built, the 'Calibration Rule' based on the velocity of material flow is researched and presented, the effect of technical and geometrical parameters to the calibration is discussed, a multi\|parameter continuous calibration is built based on single parameter discrete calibrations and neural network, and the calibration is validated through the simulation of the drawing process of a box.

关 键 词：拉延筋 标定 神经网络 板料成形 数值模拟 等效模型 

分 类 号：TG386[金属学及工艺—金属压力加工]